电信科学 ›› 2020, Vol. 36 ›› Issue (8): 112-121.doi: 10.11959/j.issn.1000-0801.2020249

• 研究与开发 • 上一篇    下一篇

基于加权随机森林算法的空巢电力用户识别方法

卢子萌1,陈佳怡2,李璟1,谢岳1,蒋欣利2,韩蕾3,郭倩1   

  1. 1 中国计量大学机电工程学院,浙江 杭州 310018
    2 国网金华供电公司,浙江 金华 321000
    3 浙江华云信息科技有限公司,浙江 杭州 310018
  • 修回日期:2020-08-05 出版日期:2020-08-20 发布日期:2020-08-26
  • 作者简介:卢子萌(1995- ),男,中国计量大学机电工程学院硕士生,主要从事电力系统大数据分析研究工作|陈佳怡(1993- ),女,国网金华供电公司工程师,主要从事营销服务与综合能源创新优化工作|李璟(1977- ),女,博士,中国计量大学机电工程学院副教授,主要从事数值分析、数据处理数据挖掘等算法研究工作|谢岳(1964- ),男,博士,中国计量大学机电工程学院教授,主要研究方向为检测技术与自动化装置|蒋欣利(1993- ),男,国网金华供电公司助理工程师,主要从事营销计量与大数据分析工作|韩蕾(1979- ),女,浙江华云信息科技有限公司中级工程师,主要从事电力营销信息化工作|郭倩(1987- ),女,博士,中国计量大学机电工程学院讲师,主要从事电力新能源分布式发电及控制技术与电力大数据分析工作
  • 基金资助:
    浙江省基础公益研究计划项目(LGG20E070003)

An empty-nest power user identification method based on weighted random forest algorithm

Zimeng LU1,Jiayi CHEN2,Jing LI1,Yue XIE1,Xinli JIANG2,Lei HAN3,Qian GUO1   

  1. 1 School of Mechanical and Electrical Engineering,China Jiliang University,Hangzhou 310018,China
    2 State Grid Jinhua Power Supply Company,Jinhua 321000,China
    3 Zhejiang Huayun Information Technology Co.,Ltd.,Hangzhou 310018,China
  • Revised:2020-08-05 Online:2020-08-20 Published:2020-08-26
  • Supported by:
    Basic Public Welfare Research Project of Zhejiang Province of China(LGG20E070003)

摘要:

针对当前政府和社会对空巢老人的识别缺乏有效技术手段的问题,提出了一种基于加权随机森林算法的空巢电力用户识别方法。首先通过调查问卷获取部分准确空巢用户标签,并从用电水平、用电波动、用电趋势 3 个方面构建用户用电特征库,由于空巢与非空巢存在用户数据不平衡问题,采用加权随机森林算法改善机器学习对数据敏感的现象,将该算法模型在电力公司采集系统部署上线,并对2 000户未知类型用户进行空巢识别,其空巢识别准确率达到 74.2%。结果表明,从用电角度研究对空巢老人的识别,可以帮助电网公司了解空巢老人的个性化、差异化需求,从而为用户提供更精细的服务,也可以协助政府和社会开展帮扶工作。

关键词: 空巢用户识别, 加权随机森林算法, 用户用电特征库, 数据不平衡

Abstract:

In view of the lack of effective technical means for the identification of empty-nesters by the government and the society,an empty-nesters prow user identification method based on weighted random forest algorithm was proposed.Firstly,some accurate labels of empty-nest users were obtained through questionnaires,and electricity characteristic library was drawn from three aspects:electricity consumption level,electricity consumption fluctuation and electricity consumption trend.Due to the data imbalance between empty-nest and non-empty-nest users,the weighted random forest algorithm was used to improve the data sensitivity phenomenon of machine learning.Finally,the algorithm model was put online in the power company’s acquisition system.The 2 000 unknown users of various types were identified,among which the identification accuracy of empty-nest users was 74.2%.The results show that the identification of empty-nesters from the perspective of electricity consumption can help power grid companies to understand the personalized and differentiated needs of empty-nesters,so as to provide users with more sophisticated services,and also assist the government and society to carry out assistance work.

Key words: empty-nest user identification, weighted random forest algorithm, user electricity characteristic library, data imbalance

中图分类号: 

No Suggested Reading articles found!